Cardiac Dysfunction and Metabolic Inflexibility in a Mouse Model of Diabetes Without Dyslipidemia

Diabetes is an increasing health burden worldwide and a major risk factor for developing cardiovascular disease and heart failure (1,2). Alterations in cardiac metabolism, including changes in substrate utilization and mitochondrial dysfunction, contribute to impaired heart function in diabetes (3–6). The heart uses both glucose and free fatty acids (FFA) as fuel, with FFA accounting for 60–70% of energy generation under normal conditions (7). FFA are metabolized to acetyl-CoA by β-oxidation, and if FFA uptake exceeds β-oxidation, as in obesity, this leads to accumulation of lipids and lipid metabolites, eventually causing lipotoxicity. By contrast, chronic hyperglycemia produces glucotoxicity, causing the formation of reactive oxygen species and advanced glycation end products (6). Impaired cardiac function is common to all types of diabetes and not confined to individuals with diabetes with obesity or dyslipidemia. However, the relative contributions of lipotoxicity and glucotoxicity to cardiac dysfunction in diabetes remain unclear.

Most studies to date have focused on dyslipidemia and shown that elevated serum FFA drive metabolic alterations in the heart (6,8–10). The role of hyperglycemia has been less well investigated. However, the fact that the hypertrophic cardiomyopathy and left ventricular dysfunction in a mouse model of lipodystrophy were ameliorated by lowering blood glucose levels (11) suggests hyperglycemia may contribute to impaired cardiac function in diabetes. Furthermore, nonobese patients with diabetes (including type 1 or monogenic diabetes) may also develop cardiac complications (12,13), supporting the idea that hyperglycemia alone is sufficient to cause cardiac disease. Likewise, in a large study of individuals with type 2 diabetes (T2D), a 1% increase in HbA_1c was associated with an increased risk of heart failure, independent of obesity, suggesting hyperglycemia is an independent risk factor for cardiac dysfunction (14).

In this study, we used an inducible mouse model of human neonatal diabetes to explore the effects of chronic hyperglycemia/hypoinsulinemia in the absence of dyslipidemia (15,16) (βV59M mice). Our mouse expresses an activating mutation (Kir6.2-V59M) in the K_ATP channel specifically in pancreatic β-cells, which prevents glucose-stimulated insulin secretion. It has several advantages over other mouse models of diabetes. First, diabetes is caused by a β-cell–specific genetic defect rather than a toxin (e.g., streptozotocin) that may have deleterious effects in other tissues. Second, diabetes is not associated with obesity or insulin resistance. Third, diabetes can be induced in adult life, precluding compensatory developmental changes. Fourth, diabetes is rapidly reversible by treatment with sulphonylurea drugs (e.g., glibenclamide), which close the open K_ATP channels (13). It should be recognized, however, that this mouse is both hyperglycemic and hypoinsulinemic, as is the case in both type 1 diabetes and nonobese T2D.

In this study, we show that plasma hyperglycemia (in conjunction with hypoinsulinemia) is a major driver of impaired cardiac metabolism and function. Gene induction caused a substrate switch from glucose oxidation toward lactate production and increased fatty acid (FA) metabolism. This led to a reduced cardiac output and stroke volume. Restoration of euglycemia by glibenclamide completely restored cardiac metabolism and function after 2 weeks of diabetes. The time-dependent deterioration of cardiac function correlated with reduced glucose flux through pyruvate dehydrogenase (PDH) and increased expression of PDH kinase 4 (PDK4) and uncoupling protein 3 (UCP3).

Animal studies were conducted in accordance with the U.K. Animals (Scientific Procedures) Act (1986) and local ethical guidelines (Medical Research Council’s Responsibility in the Use of Animals in Medical Research, 1993). Mice hemizygously expressing an inducible Kir6.2-V59M transgene selectively in pancreatic β-cells (βV59M mice) were generated and transgene expression induced by tamoxifen injection, as described (15). In some experiments, βV59M mice were used before gene induction as their own controls. In other experiments, tamoxifen-injected or uninjected wild-type, RIPII-Cre-ER, and floxed Kir6.2-V59M gene littermates were used as controls. Mice were maintained on a 12-h light/dark cycle at 21 ± 2°C with an unrestricted diet (63% carbohydrate, 23% protein, and 4% fat; RM3, Special Diet Services).

Body weight and blood glucose levels were monitored routinely. For glibenclamide treatment, after 2 weeks of diabetes, mice were subcutaneously implanted with two pellets, each releasing 25 mg glibenclamide over 21 days (Innovative Research of America) and under 2% isoflurane anesthesia. A schematic overview of the mouse experiments is given in Supplementary Fig. 1D. For tissue analysis, mice were culled by cervical dislocation, and hearts were collected, snap-frozen in liquid nitrogen, pulverized, and stored at −80°C for later analysis.

Blood glucose levels were measured from the tail vein using a FreeStyle Lite device and FreeStyle Lite test strips (Abbott Laboratories). Serum was obtained by incubating whole blood on ice for 30 min followed by centrifugation at 3,000g and 4°C for 30 min. It was then snap-frozen in liquid nitrogen for later analysis. Serum glucose was measured using a glucose (HK) kit (Sigma-Aldrich). Serum triglyceride (ab65336; Abcam), FFA (ab65341; Abcam), cholesterol (ab65390; Abcam), and insulin (10-1247-01; Mercodia) levels were measured using the indicated kits.

Mice were imaged on an 11.7T MRI instrument (Bruker) as previously described (17). Eight to 10 short-axis slices (slice thickness, 1.0 mm; matrix size, 256 × 256; field of view, 25.6 × 25.6 mm; echo time/repetition time, 1.43/4.6 ms; flip angle, 17.5°; and number of averages, 4) were acquired with a cine fast low-angle shot sequence (18). Left ventricular volumes were derived using the freehand draw function in ImageJ (National Institutes of Health). For each heart, left ventricular mass, ejection fraction, stroke volume, and cardiac output were calculated.

Experiments were performed between 7 and 11 a.m. when mice were in the fed state. A total of 40 mg [1-¹³C]pyruvate (Sigma-Aldrich) doped with 15 mmol/L trityl radical (OXO63; GE Healthcare) and 3 μL Dotarem (1:50 dilution; Guerbet) was hyperpolarized in a prototype polarizer, with 20–30 min of microwave irradiation (19). The sample was subsequently dissolved in a pressurized and heated alkaline solution, containing 2.4 g/L sodium hydroxide and 100 mg/L EDTA dipotassium salt (Sigma-Aldrich), to yield a solution of 80 mmol/L hyperpolarized sodium [1-¹³C]pyruvate with a polarization of 30%. A total of 200 μL was injected over 10 s via the tail vein. ¹³C MR pulse-acquire cardiac spectra were acquired over 60 s following injection of hyperpolarized [1-¹³C]pyruvate (repetition time, 1 s; excitation flip angle, 15°; sweep width, 13,021 Hz; acquired points, 2,048; and frequency centered on the C₁ pyruvate resonance) (20). The ¹³C label from pyruvate and its metabolic products was summed over 30 s from the first appearance of pyruvate and fitted with the AMARES algorithm in jMRUI (21). Data are reported as the ratio of metabolic product to the [1-¹³C]pyruvate signal to normalize for differences in polarization and delivery.

Cardiac triglycerides were isolated using the chloroform/methanol extraction (adapted from Folch et al. [22]). In brief, frozen, pulverized tissue was lysed in chloroform/methanol (2:1) using a Tissue Lyser (Qiagen) at 30 Hz for 1 min, incubated for 20 min at 20°C with vigorous shaking (1,400 rpm; ThermoMixer; Eppendorf), and centrifuged at 13,000 rpm for 30 min at 20°C. The liquid phase was mixed with 0.9% NaCl and centrifuged at 2,000 rpm for 5 min. Chloroform/Triton X-100 (1:1) was added to the organic phase and solvent evaporated under the fume hood. Triglycerides were measured using the ab65336 kit (Abcam) and values normalized to the tissue pellet weight.

Glycogen was extracted from frozen, pulverized tissue samples using 30% KOH and a Tissue Lyser (30 Hz for 1 min; Qiagen), followed by incubation at 95°C for 30 min. Glycogen was precipitated by addition of 95% ethanol and centrifugation at 3,000g for 20 min. Pellets were dissolved in water and digested with amyloglucosidase (Sigma-Aldrich). Glucose was measured using a glucose (HK) kit (Sigma-Aldrich) and normalized to protein content (Pierce BCA protein assay; Thermo Fisher Scientific).

Proteins were extracted from frozen, pulverized tissue samples in a Tissue Lyser (30 Hz for 1 min; Qiagen) using a protein lysis buffer containing 50 mmol/L Tris, 150 mmol/L NaCl, 1 mmol/L EDTA, 10 mmol/L NaF, 2 mmol/L Na₃VO₄, 1 mmol/L dithiothreitol, 1× protease inhibitor cocktail (Sigma-Aldrich), and 1% Nonidet P-40. Protein concentration was measured using a Pierce BCA protein assay kit (Thermo Fisher Scientific). Proteins were separated on 4–12% Bis-Tris SDS-PAGE gels (NuPAGE Novex; Invitrogen), transferred to nitrocellulose membranes, and detected using specific antibodies to PDK4 (8), medium-chain acyl-CoA dehydrogenase (MCAD; sc-49047; Santa Cruz Biotechnology), UCP3 (ab3477; Abcam), HSC70 (ab19136; Abcam), and vinculin (ab129002; Abcam). For assessment of protein oxidation, we used an Oxidized Protein Western Blot kit (ab178020; Abcam). Proteins were quantified using ImageJ (National Institutes of Health).

Pulverized, frozen heart tissue was lysed in water, extracted, and processed as recommended using a Protein Carbonyl Content Assay kit (ab126287; Abcam).

Samples were prepared for liquid chromatography–tandem mass spectrometry (LC-MS/MS) analysis as described (23). Briefly, proteins were precipitated with chloroform/methanol following cell lysis in radioimmunoprecipitation assay buffer and reduction/alkylation with dithiothreitol/iodoacetamide. After protein digestion with trypsin, peptides were purified on reverse-phase material (SOLA SPE; Thermo Fisher Scientific) and injected into a nano–LC-MS/MS workflow consisting of a Dionex Ultimate 3000 UPLC and an Orbitrap Fusion Lumos instrument (both from Thermo Fisher Scientific). Peptides were separated on an easy-spray column (500 mm × 75 µm) with a flow rate of 250 nL/min and a gradient of 2–35% acetonitrile in 5% DMSO/0.1% formic acid within 60 min. Detailed MS instrument settings are listed in Supplementary Table 1.

LC-MS/MS data were analyzed using label-free precursor quantitation in Progenesis QI (version 3.0.6039.34628; Waters) and peptides identified with Mascot v2.7 (Matrix Science) against the UniProt/Swiss-Prot database (retrieved 26 November 2015). Peptide false discovery rate was adjusted to 1%, and additionally, all spectra identified with a score <20 were discarded.

Mice were infused with two boluses of 20% U-¹³C-glucose (Cambridge Isotope Laboratories) in saline by intraperitoneal injection at 2 mg U-¹³C-glucose/g mouse weight at 0 and 15 min. Hearts were harvested at 30 min post–first bolus and immediately quenched by freeze-clamping at −80°C. Frozen tissue was lyophilized and pulverized. Powdered material was transferred to a 1.5-mL Eppendorf tube and masses recorded. A total of 600 μL chloroform/methanol (2:1 volume for volume) was added to each sample and vortexed briefly before pulse sonication (3 × 8 min) in a water-bath sonicator at 4°C for 1 h. Samples were spun (13,200 rpm, 4°C, 10 min), supernatant transferred to a new tube, and dried in a rotary vacuum concentrator. The remaining pellet was re-extracted with 600 μL methanol/water (2:1 volume for volume, containing 25 nmol nor-Leucine [internal standard 1]) followed by pulse sonication for 8 min at 4°C. Samples were spun (as above) and the supernatant added to the first extract and dried. Extracts were resuspended in 350 μL chloroform/methanol/water (1:3:3) to partition polar and apolar metabolites. A total of X μL of aqueous (polar) phase was dried with 1 nmol scyllo-inositol (internal standard 2, X = volume of extract equivalent to 1 mg dry weight of heart tissue). Samples were washed twice with methanol and derivatized with 20 μL methoxyamine HCl (20 mg/mL in pyridine, overnight, at room temperature) and 20 μL BSTFA + TMCS (Thermo Fisher Scientific) for >1 h.

Metabolite analysis was performed using gas chromatography-MS (GC-MS) using a 7890B-5977A system (Agilent Technologies). Splitless injection (injection temperature 270°C) onto a 30 m + 10 m × 0.25 mm DB-5MS+DG column (Agilent Technologies) was used, with helium as the carrier gas, in electron impact ionization mode. The initial oven temperature was 70°C (2 min), followed by temperature gradients to 295°C at 12.5°C/min and then to 320°C at 25°C/min (held for 3 min). Metabolites were identified and quantified by comparison with the retention times and mass spectra of authentic standards using MassHunter WorkStation software (B.06.00 SP01; Agilent Technologies). Label incorporation and abundance was estimated using GAVIN software (24). Label incorporation was insignificant in diabetic mice due to their high level of (unlabeled) blood glucose.

DNA was extracted from frozen, pulverized tissue samples using a DNeasy Blood and Tissue kit (Qiagen) following the manufacturer’s instructions.

RNA was extracted from frozen, pulverized tissue samples using Qiazol lysis reagent and a Tissue Lyser (30 Hz, 1 min), followed by an RNeasy Mini kit (all from Qiagen). For determination of expression of mitochondrially encoded genes, we included an on-column DNase digestion step (Qiagen). cDNA was transcribed using the High Capacity cDNA Reverse Transcription kit (Applied Biosystems). Quantitative PCRs were performed using TaqMan probes (Supplementary Table 2) and the fast reaction kit on the StepOnePlus instrument (Life Technologies). Data were quantified according to the delta threshold cycle method (25) and normalized to levels of actin RNA.

For determination of genomic and mitochondrial DNA, we performed a semiquantitative SYBR PCR using the Power SYBR Green PCR Master Mix (Applied Biosystems) and the following primers: mtCox2 forward, 5′-ATAACCGAGTCGTTCTGCCAAT-3′ and mtCox2 reverse, 5′-TTTCAGAGCATTGGCCATAGAA-3′; and Rsp18 forward, 5′-TGTGTTAGGGGACTGGTGGACA-3′ and Rsp18 reverse, 5′-CATCACCCACTTACCCCCAAAA-3′.

Data are given as mean ± SEM of the indicated number of mice (n). Significance was tested by t test or one-way or two-way ANOVA as indicated. Where data were found not to be normally distributed, nonparametric tests were used (i.e., Kruskal-Wallis test). Differences between groups were considered statistically significant if P was <0.05.

Transgene induction in adult βV59M mice (15,16) led to marked hyperglycemia (blood glucose >20 mmol/L) within 24 h, and serum glucose remained elevated over the next 12 weeks (Supplementary Fig. 1A and C). Serum insulin levels were significantly reduced after 2 weeks and remained so over the next 20 weeks of diabetes (Fig. 1A). Cardiac glycogen levels were unaltered (Supplementary Fig. 1B). We detected no changes in serum triglyceride, FA, or cholesterol levels or lipid accumulation in the liver or heart after 2 weeks of diabetes (Table 1 and Fig. 1B). Serum triglyceride levels were slightly elevated after 4 and 20 weeks of diabetes, hence we focused on studying 2-week diabetic mice.

Glibenclamide therapy (15) reversed the hyperglycemia within 2 days (Supplementary Fig. 1C). Two weeks later, free-fed and fasted plasma glucose and serum insulin levels were normal. Serum and organ lipids remained unaltered (Fig. 1A and B and Table 1).

No significant change in cardiac size was seen with 2 weeks of diabetes (Supplementary Fig. 2A–D), but there was a significant decrease in both stroke volume and cardiac output (Fig. 1C and D). This was associated with a marked change in cardiac metabolism, as measured by both hyperpolarized MRS (HP-MRS) (Fig. 1E) and metabolomics (Fig. 2).

Bicarbonate production from hyperpolarized [1-¹³C]pyruvate decreased in hearts of diabetic mice (Fig. 1E), indicating reduced pyruvate flux though PDH (9). Neither lactate nor alanine production from [1-¹³C]pyruvate were significantly altered, although there was a trend toward an increase in lactate (Supplementary Fig. 2E and F). Glibenclamide therapy (2 weeks) completely reversed the impairment of cardiac function and pyruvate metabolism induced by 2 weeks of hyperglycemia (Fig. 1C–E and Supplementary Fig. 2).

Diabetes also affected metabolite abundance in the heart when measured in mice injected with U-¹³C-glucose using GC-MS. There was a significant increase in pyruvate abundance (Fig. 2A), indicating either enhanced pyruvate production or reduced consumption. In light of the HP-MRS data (Fig. 1E), the latter is the more likely. Lactate and alanine were also increased (Fig. 2A), again indicative of glycolytic flux being diverted from the tricarboxyclic acid (TCA) cycle into lactate production. As expected, glucose was increased, and there was a trend toward an increase in other glycolytic intermediates (Fig. 2C). Fructose was strongly elevated (Fig. 2D), as were the branched-chain amino acids (BCAA) isoleucine and valine (Fig. 2E). Somewhat counterintuitively, and despite the unaltered abundance of many other metabolites (Supplementary Fig. 3), the TCA cycle metabolites citrate, cis-aconitate, succinate, and fumarate were significantly increased (Fig. 2B). This may reflect a relative reduction in TCA turnover as a consequence of decreased pyruvate flux through PDH, resulting in “pooling” of these metabolites. Alternatively, the TCA cycle may switch to being fueled by FA β-oxidation (for acetyl-CoA) and/or glutaminolysis (for carbon skeletons). In support of this, hyperglycemia also reduced the relative abundance of some (but not all) longer chain FA, including C14:0, C16:1, and C18:1a (Fig. 2F and Supplementary Fig. 3). Taken together, these data argue that chronic hyperglycemia/hypoinsulinemia induces a metabolic switch from oxidative metabolism of glucose to lactate production and a concomitant increase in FA β-oxidation.

Ideally, to assess fluxes, it would be valuable to measure the amount of each metabolite that derived from ¹³C-labeled glucose. This was possible in wild-type animals but not in diabetic mice because their high blood glucose levels meant that the percentage of labeled glucose was too small to be resolved.

We next explored if the switch in cardiac metabolism induced by diabetes was due to changes in the levels of metabolic proteins by performing global protein expression profiling of hearts from control mice and mice that had been diabetic for 24 h or 2 weeks (Fig. 3). A total of 2,449 grouped proteins at 1% false discovery rate were identified (2,228 quantified). Of these, 299 proteins exhibited differential expression between at least two conditions (P < 0.05 by ANOVA) (Supplementary Table 3).

Consistent with the HP-MRS and metabolomics data, diabetes increased the abundance of proteins involved in FA metabolism and altered the abundance of many proteins involved in glucose oxidation (Fig. 3 and Supplementary Table 3).

PDK4 was the most strongly upregulated protein in the glucose oxidation protein cluster (4.8-fold) (Fig. 3). PDK4 regulates the activity of PDH by phosphorylating and inactivating the enzyme. Its upregulation may explain why PDH flux is reduced in diabetes. Two key enzymes in the TCA cycle, citrate synthase and oxoglutarate dehydrogenase (OGDH), were downregulated. The latter is of special significance as it is the rate-limiting enzyme of the TCA cycle, and its downregulation will reduce TCA cycle activity. The defect in TCA cycle activity was not associated with altered mitochondrial number (as measured by mitochondrial vs. genomic DNA content and expression of mitochondrially encoded genes) (Supplementary Fig. 4A and B) nor with enhanced protein oxidation/carbonylation (Supplementary Fig. 4C and D). Expression of several enzymes involved in the late steps of glycolysis, including phosphoglycerate kinase 1, β-enolase (ENO3), and pyruvate kinase, was downregulated. Lactate dehydrogenase (LDHB) and the lactate transporter (MCT1) were also reduced, despite the elevated lactate production.

Notwithstanding the fact that the heart is not considered to be a ketogenic organ, the most strongly elevated protein involved in lipid metabolism was 3-hydroxy-3-methylglutaryl-CoA synthase 2 (mitochondrial, 21.4-fold), which catalyzes the first step in ketogenesis. The mitochondrial carnitine/acylcarnitine carrier protein SLC25A20, which mediates FA import into mitochondria for oxidation, was also increased (2.1-fold). In addition, many proteins involved in FA oxidation (e.g., ACOX1 and SCP2) or lipolysis (e.g., PLIN5 and MGL) were elevated (Fig. 3).

Interestingly, proteins involved in contractile function (such as myoglobin, α-actinin 2, and myosin 6) were downregulated in diabetic hearts (Supplementary Table 3), as were proteins involved in the creatine kinase shuttle (e.g., CKMT2 and CKB).

In summary, the proteomics data support the metabolic data in suggesting glucose oxidation is impaired and FA metabolism enhanced.

In line with the proteomics data, diabetes dramatically increased PDK4 mRNA (Fig. 4A) and protein expression (Fig. 4B). PDK4 is transcriptionally regulated by peroxisome proliferator–activated receptor α (PPARα), which is upregulated in the hearts of many rodent models of diabetes (26–28). However, Pparα mRNA expression in βV59M mouse heart was unchanged by hyperglycemia (Fig. 4C). Similarly, Pgc1α mRNA, which is elevated in other mouse models of diabetes (26,27) and regulates both PDK4 expression and PDH activity (29), was unaltered (Fig. 4D). Expression of the GLUT Glut4 decreased in βV59M diabetic mouse hearts (Fig. 4E).

MCAD and UCP3 are major regulators of FA usage in the heart (9,30), and their expression increases in hearts of diabetic rodents with dyslipidemia (9). However, MCAD mRNA and protein were unaltered by diabetes in βV59M hearts (Fig. 5A and C and Supplementary Fig. 4E). This is consistent with the unchanged levels of cardiac lipids (Table 1) and Pparα/Pgc1α mRNA expression. Although PPARα is also regulated posttranscriptionally by ligand (lipid) binding, the lack of an increase in expression of its target gene (MCAD) suggests PPARα is not a major transcriptional regulator in this setting. Both UCP3 mRNA and protein increased significantly (Fig. 5B and C and Supplementary Fig. 4).

The increase in Ucp3 and Pdk4 mRNA was observed as early as 24 h after gene induction, and elevated levels persisted over the next 20 weeks (Fig. 6A and B). At the protein level, increased amounts of PDK4 and UCP3 were only detectable after 2 and 4 weeks of diabetes, respectively (Fig. 6C). However, a reduction in cardiac output and stroke volume was observed within 2 weeks of diabetes induction (Figs. 1C and D and 6D and E). Prolonged diabetes (8 weeks) led to further deterioration of cardiac function, with both the ejection fraction and heart rate also being reduced (Fig. 6F and G).

Using a combination of HP-MRS, metabolomics, MRI, and proteomics, we show that diabetes (hyperglycemia/hypoinsulinemia) drives changes in cardiac metabolism independent of (or in addition to) plasma FA/lipids. In particular, it reduces flux through PDH and thereby impairs entry of pyruvate into the TCA cycle. Although mitochondrial numbers seem unaltered, the relative amount of glucose oxidized by the mitochondria is reduced, leading to the observed increase in lactate production. Proteins involved in FA metabolism increase, and long-chain FA levels decrease, indicating FA metabolism is enhanced. These alterations in cardiac metabolism lead to impaired cardiac function, as indicated by the reduction in cardiac output and stroke volume.

We found that changes in cardiac metabolism manifest very rapidly: PDK4 expression is already altered within 24 h of diabetes induction. Thus, glucose is a very fast driver of cardiac dysfunction. Indeed, although some studies found increased PDK4 levels already after 24 h of high-fat diet feeding (31), others show that high-fat feeding only impairs cardiac metabolism and function after a longer exposure (30). We also demonstrate that the changes in cardiac metabolism and function induced by diabetes are reversed when euglycemia is restored.

Importantly, because diabetic βV59M mice are not obese and show no changes in serum or tissue lipids at the time point studied, the effects we observe are the result of hyperglycemia/hypoinsulinemia and not obesity or dyslipidemia. Our findings are supported by a recent study of the Bscl2^−/− mouse model of lipodystrophy (11). These mice exhibited cardiac dysfunction without any change in intramyocardial lipids, which was suggested to result from elevated plasma glucose, as it could be reversed by the glucose-lowering sodium–glucose cotransporter 2 inhibitor dapagliflozin.

The hyperglycemia of βV59M mice results from hypoinsulinemia. Several lines of evidence suggest that it is primarily hyperglycemia and not hypoinsulinemia that drives the cardiac changes we observe. First, serum insulin is reduced but not absent in βV59M mice (15) (Fig. 1A). Second, preventing hyperglycemia using dapagliflozin preserved cardiac function in a mouse model of lipodystrophy (11). Third, deletion of the insulin receptor specifically in cardiac myocytes (preventing insulin action) produced very different changes in gene expression and metabolism to those seen in βV59M mice (32). For example, PDK4 and MCAD expression were downregulated, GLUT4 expression was upregulated, and FA oxidation was reduced; these changes are all the opposite of those seen in our mice. Nevertheless, although we favor the idea that chronic hyperglycemia is the primary cause of many of the changes we observe, we cannot exclude a role for hypoinsulinemia. Diabetes is a multifactorial disease, and both hyperglycemia and hypoinsulinemia are important. We can, however, exclude a role for serum dyslipidemia in our model.

As previously reported for other diabetes models (33), diabetic βV59M mice had lower expression of Glut4 in the heart. Nevertheless, cytosolic glucose levels were slightly elevated in diabetes, presumably due to the marked increase in blood glucose. This suggests the decrease in GLUT4 is a protective mechanism, producing a relative reduction in glucose uptake in the face of chronic hyperglycemia.

Many proteins involved in glucose metabolism, such as pyruvate kinase, citrate synthase, and OGDH, were decreased in diabetes. OGDH is a subunit of the enzyme complex that converts 2-oxoglutarate to succinyl-CoA and CO₂. This is the rate-limiting reaction in the TCA cycle (34), and a reduction in OGDH, together with the reduced PDH flux, may lead to impaired ATP generation by diabetic hearts, causing functional defects. Mitochondrial dysfunction and reduced glucose oxidation have been reported in mitochondria isolated from hearts of donors with T2D (4).

It was somewhat surprising that we saw no increase in protein carbonylation in diabetic hearts, as it is well established that cardiac oxidative stress occurs in diabetes (35). The extent of carbonylation is considered a marker for oxidative stress, with the advantage that it is relatively stable and induced by almost all types of reactive oxygen species (36). It is possible that our failure to detect protein carbonylation indicates that 2–4 weeks of diabetes is insufficient to induce measurable oxidative stress, it is insufficient for oxidative stress to produce protein carbonylation, or oxidative stress is less when hyperglycemia/hypoinsulinemia is not accompanied by dyslipidemia. It may, however, explain why we observe a near-complete reversibility of the cardiac phenotype upon restoration of euglycemia.

Many proteins involved in lipid metabolism were upregulated in diabetes. These included proteins involved in ketogenesis (e.g., 3-hydroxy-3-methylglutaryl-CoA synthase 2), β-oxidation (ACOX1 and SCP2), mitochondrial FA import (SLC25A20), and lipolysis (PLIN5 and MGL). These changes suggest enhanced β-oxidation in hyperglycemia. Metabolomics analysis supports this idea, revealing a marked decrease in myristic (C14:0) and palmitoleic (C16:1) acids. Thus, our data suggest that hyperglycemia/hypoinsulinemia leads to a paradoxical increase in FA usage and a concomitant decrease in glucose usage in the diabetic heart, perhaps contributing to TCA flux maintenance.

Metabolomics analysis demonstrated a marked increase in the BCAA isoleucine and valine in diabetic hearts. It is possible this reflects plasma levels of these substances, which are enhanced in diabetes (37). However, the heart itself catabolizes BCAA and expresses high levels of enzymes involved in BCAA oxidation (38), such as BCAA aminotransferase and branched-chain keto acid dehydrogenase (BCKDHA and BCKDHB). These enzymes are also highly expressed in the mouse heart, as detected by proteomics, but they are not significantly altered by diabetes (Supplementary Table 3). It is also noteworthy that inhibition of cardiac BCAA catabolism in mice leads to decreased PDH activity and glucose oxidation (39).

PDH is the key enzyme regulating substrate usage in the heart. Reduced PDH activity decreases mitochondrial carbohydrate metabolism and enhances lactate production. This favors FA usage, impairing cardiac energy production and thereby cardiac function (40). Cardiac PDH activity is decreased in many animal models of diabetes (9,41,42), most of which are both hyperglycemic and dyslipidemic, and cardiac Pdk4 mRNA levels are dramatically increased in streptozotocin-injected diabetic rats (43). As we show in this study, hyperglycemia/hypoinsulinemia is sufficient to reduce PDH activity, likely via transcriptional upregulation of PDK4. This does not exclude a role for FA in regulating PDK4 expression. Indeed, PDK4 expression is induced by long-chain FA (e.g., palmitate) (44), and both increased Pdk4 mRNA and reduced PDH flux were found in hearts of rats fed a high-fat diet (9). Thus, an additional contribution from elevated lipids may occur when hyperglycemia is associated with obesity and dyslipidemia. Interestingly, inhibition of PDK activity by the pyruvate mimetic dichloroacetate normalized cardiac metabolism and diastolic function in high-fat–fed rats, arguing that reduced PDH flux contributes to diabetes-associated cardiac dysfunction (9).

The major transcription factor regulating PDK4 in the heart is thought to be PPARα (9). It also elevates MCAD and UCP3 expression (9). Although PPARα expression is induced in diabetes (26–28), PPARα is primarily regulated posttranscriptionally (e.g., by lipids [45]). There was little change in intracellular lipids and Pparα mRNA expression in diabetic βV59M hearts. Furthermore, the lack of a change in MCAD expression argues PPARα activity may also be unchanged and that other factors may underlie the marked increase in PDK4 expression. In this context, it is important to recall that PPAR response elements are absent in the promoter region of PDK4 (46).

Our results demonstrate that hyperglycemia/hypoinsulinemia alone is sufficient to impair cardiac function, that this can occur very rapidly following elevation of blood glucose, and that (at least after diabetes of short duration) it may be reversed by restoration of euglycemia. This suggests that even nonobese patients with diabetes, or those without systemic hyperlipidemia, may already have some cardiac impairment at presentation. It also emphasizes the need for tight control of blood glucose levels, which may not only prevent but also ameliorate the deleterious effects of diabetes on the heart.

Acknowledgments. The authors thank Lisa Heather, Duncan Sparrow, Mike Dodd (Department of Physiology, Anatomy and Genetics, University of Oxford), and Markus Ralser (The Francis Crick Institute, London) for helpful discussion and Raul Terron Exposito, Olof Rorsman, and Idoia Portillo (Department of Physiology, Anatomy and Genetics, University of Oxford) for technical support. Proteomics analysis was performed in the Target Discovery Institute Mass Spectrometry Laboratory led by Benedikt M. Kessler.

Funding. This work was supported by the Wellcome Trust (grants 084655 and 089795) and the European Research Council (ERC advanced grant 322620). M.R. was supported by a Novo Nordisk postdoctoral fellowship run in partnership with the University of Oxford. R.F. was supported by the Kennedy Trust Fund. Metabolomics was supported by The Francis Crick Institute, which receives its core funding from Cancer Research UK (FC001999), the Medical Research Council (FC001999), and the Wellcome Trust (FC001999). F.M.A. held a Royal Society/Wolfson Merit Award and a European Research Council Advanced Investigatorship. The HP-MRS studies were supported by the British Heart Foundation (Fellowships FS/10/002/28078 and FS/14/17/30634), the Oxford British Heart Foundation Centre of Research Excellence (grant RE/13/1/30181), and the Danish Council for Strategic Research (LIFE-DNP Programme).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. M.R. performed the animal and molecular biology work and analyzed the data. M.R. and F.M.A. designed the study and wrote the manuscript. D.S., V.B., and M.K.C. performed the cine MRI and HP-MRS measurements. D.S. and D.J.T. analyzed the cine MRI and HP-MRS data. S.B. and R.F. performed proteomics and analyzed the data. N.L. and J.I.M. performed metabolomics and analyzed the data. All authors read the manuscript and discussed the interpretation of the results. D.J.T. and F.M.A. are the guarantors of this work and, as such, had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.